SMOC

From the Mercator project to Copernicus Marine

In 1995 a group of some thirty French scientists launched the Mercator project, named after the 16^th-century Flemish cartographer, to describe and forecast the ocean the way weather services forecast the atmosphere. It was carried by five French institutions working at sea: CNRS, Ifremer, IRD, Météo-France and SHOM, the French navy's hydrographic and oceanographic service. After six years of prototypes the first Mercator bulletin came out in 2001, mapping currents, temperature and salinity over the North and Equatorial Atlantic.

Mercator Ocean led the European MyOcean projects from 2009, and on 11 November 2014 the European Union entrusted it with the Copernicus Marine Environment Monitoring Service, which opened as an operational service in May 2015. A global configuration at 1/12° had been running in real time since 2011; its version PSY4V3, since renamed GLO12v3, took over on 19 October 2016, and the present GLO12v4 has been in service since November 2022. The organisation is now becoming intergovernmental: in June 2025, at the UN Ocean Conference in Nice, twelve European countries backed the creation of the Mercator International Centre for the Ocean.

SMOC, for Surface and Merged Ocean Currents, joined the global product in January 2019. It was built for drift: what moves a floating object is not only the large-scale circulation but the whole velocity felt at the surface, tide and waves included. Its forecast was stretched to 10 days in May 2021, the same range as the ocean model it is built on.

An ocean model, a tide model and a wave model

SMOC is not one model but the sum of three. Its first and main part, the ocean circulation, comes from GLO12v4, built on NEMO 3.6 (Nucleus for European Modelling of the Ocean). NEMO solves the primitive equations of a hydrostatic ocean under the Boussinesq approximation, which neglects changes of density except where they make water rise or sink. It is coupled with the LIM3 sea ice model and its 11 categories of ice.

The model runs on ORCA12, a tripolar grid whose northern poles are moved onto land so that no point of the Arctic Ocean is a singularity. Its mesh is 9 km at the equator, 7 km at mid-latitudes (off Cape Hatteras) and down to 2 km near Antarctica; the outputs are interpolated onto a regular 1/12° grid of 4320 × 2041 points. On the vertical it has 50 levels at fixed depths, from 1 m thick at the surface to 450 m at the bottom, 22 of them in the top 100 m, with partial cells to follow the sea floor.

At 1/12° the model is called eddy-resolving: the ocean's eddies and the meanders of the great boundary currents, the Gulf Stream, the Kuroshio or the Agulhas, are computed rather than smoothed into an average. Vertical mixing comes from a two-equation turbulence scheme (k-epsilon), and the model carries no tide of its own. The atmosphere drives it through ECMWF's IFS forecasts at 1/10° and hourly, turned into wind stress and heat and water fluxes by bulk formulas that take the surface current itself into account; the 2026 manual adds a wave forcing from MFWAM, and a hundred major rivers plus the melt of Greenland and Antarctica feed it fresh water.

The tide comes from FES2014, a finite-element tide model built by LEGOS, NOVELTIS and CLS for CNES, which solves the tidal equations for the whole water column and assimilates years of altimetry (Topex/Poseidon, Jason-1 and 2, ERS-1 and 2, Envisat) and tide gauges, over 34 tidal constituents. SMOC computes from it, every hour on its 1/12° grid, the barotropic tidal current, that is the depth-averaged one. The waves come from MFWAM, Météo-France's wave model for Copernicus Marine, built on ECMWF's ECWAM code and driven by IFS winds: its Stokes drift, the small net forward motion of water under passing waves, is interpolated from its 1/10° grid onto the 1/12° one.

The three parts are simply added: total current = circulation + Stokes drift + tide. The interactions between them, waves on a current or tide on the circulation, are therefore not represented, as the producers point out. And the circulation is the model's current at 0.5 m below the surface, not extrapolated to the surface itself.

A weekly analysis, a daily forecast

The ocean model is set on observations from space and from the sea: the along-track sea level anomalies of every altimetry satellite that Copernicus Marine collects, satellite sea surface temperature (ODYSSEA), sea ice concentration (OSI SAF) and in situ temperature and salinity profiles from the Coriolis database, where the Argo floats' profiles land. Sea level from altimetry is read against a mean dynamic topography (CNES-CLS18), the average relief of the sea surface that the steady currents shape. Before use, the observations falling in one model cell are merged into a single super-observation, to filter out scales the model cannot resolve.

The assimilation scheme, SAM2, is a reduced-order Kalman filter derived from SEEK (Singular Evolutive Extended Kalman). Rather than a full map of how forecast errors relate to each other, it describes them with a few hundred ocean states taken from the GLORYS12 reanalysis, and it weighs each observation against the model accordingly. It works on a 7-day window, with a 4D extension that accounts for when each observation was made, and its corrections are fed into the model gradually over the week (Incremental Analysis Update) so as not to shock it. A 3D-Var scheme over the last month adds a correction of the slow, large-scale biases in temperature and salinity.

The analysis is weekly, the forecast daily. Each day the system adds one day of simulation driven by the analysed atmosphere, then runs a new 10-day forecast from 00 UTC, delivered between 08 and 12 UTC. Each week the past simulations are replaced by analyses, and two weeks later by the final best analyses, which is why the product's archive keeps improving after the fact.

SMOC is then assembled hour by hour: the hourly surface current of GLO12, the Stokes drift of the global wave forecast and the tidal current computed from FES2014, each delivered separately and summed into a total current. Everything is published on the Copernicus Marine Data Store with a two-year rolling archive, free of charge for registered users.

At sea

A knot of current gives or takes a mile every hour, and in the open ocean SMOC shows where the big ones run: the Gulf Stream, whose surface core reaches about 9 km/h (close to 5 knots) according to NOAA, the Kuroshio, the Agulhas, the equatorial currents and the Antarctic Circumpolar Current. Altimetry keeps its eddies roughly where the satellites see them, and particles tracked in the model reproduce the large-scale paths of real drifting buoys well.

It also tends to undersell them. Against drifting buoys, the producers find mid-latitude surface currents too weak by about 20% in strong currents and up to 60% in weak ones, the strong western boundary currents being the best rendered. An eddy shifted a little from its true place, between two satellite passes, can turn a fair current into a foul one: the forecast current is a probable picture, not a certainty.

Its tide is a strength over open continental shelves, such as the approaches of the English Channel or the Celtic Sea, where the broad tidal streams turn with the hours. It is also its limit near the coast: a depth-averaged tide on a 9 km mesh cannot draw the race off a headland, the eddy behind an island or the stream in a narrow pass. In the producers' tests, adding tide and waves shortens the gap between modelled and real drifters over three days worldwide, from 42.7 to 39.9 km, but slightly lengthens it over shelf seas, from 36.4 to 38.5 km; close to land, tidal atlases and pilot books remain the reference.

The Stokes drift goes mostly with the waves, hence downwind in a blow, and fades quickly with depth: SMOC was designed for what floats at the very surface, while a keel reaches below. Next to RTOFS, NOAA's current, SMOC is a second opinion from an independent system, with another model, other assimilation and other winds: where both agree on a current, it is worth trusting; where they place an eddy differently, the current there is uncertain; near the coast most of their difference is the tide, which RTOFS leaves out.

SMOC in Saily WR

Saily WR takes SMOC's total surface current, the sum of ocean circulation, tide and wave Stokes drift that Copernicus Marine distributes, hour by hour on its 1/12° grid (about 9 km) over the whole globe, from each day's run and over its 10 days. When the sailor chooses SMOC as the current, rather than RTOFS or no current at all, it is added to the boat's speed at each point and hour of the route, on the route of every wind model, so that they stay comparable side by side.

A server downloads each new run as soon as it is published, and the Data page shows live which run SMOC is on, its horizon and its coverage. Computing the same trip with SMOC, then with RTOFS or without current, shows how much of a route's gain comes from the current, and whether it hangs on a tidal gate or an eddy.

As the Copernicus Marine licence requires, the data are credited wherever they are shown: "E.U. Copernicus Marine Service Information (CMEMS)". That licence is free and lets anyone build and share products derived from the data, for any purpose, on that one condition.